Elsevier

Neurocomputing

Volume 73, Issues 7–9, March 2010, Pages 1451-1456
Neurocomputing

Power quality disturbance classification using Hilbert transform and RBF networks

https://doi.org/10.1016/j.neucom.2009.11.008Get rights and content

Abstract

This paper presents the application of Hilbert transform and artificial neural network (ANN) for power quality (PQ) disturbance classification. The input features of the ANN are extracted from the envelope of the disturbance signals by applying Hilbert transform (HT). The features obtained from the Hilbert transform are distinct, understandable and immune to noise. These features after normalization are given to the radial basis function (RBF) neural network. The data required to develop the network are generated by simulating various faults in a test system. The performance of the proposed method is compared with the existing feature extraction techniques in combination with other ANN architectures. Simulation results show the effectiveness of the proposed method for power quality disturbance classification.

Introduction

The quality of electric power has become an important issue for electric power utilities and its customers. As a result, power quality (PQ) research is gaining interest. Degradation in quality of electric power is normally caused by power-line disturbances such as voltage sag, swell, momentary interruption, flicker, notch, transients and harmonics. These disturbances results in malfunctions, reduced life time and failure of electrical equipment. In order to determine the sources and causes of power quality disturbances, we must be able to detect and classify the disturbances into different types.

The major requirement in power quality study is the ability to perform automatic power quality monitoring and data analysis. Feature extraction is a vital step in automatic disturbance waveform classification. Spectral analysis using discrete Fourier transform (DFT) and fast Fourier transform (FFT) [1] have been applied for this purpose, but due to the non-stationary nature of the power quality disturbances such transforms are not effective in detecting the disturbance waveforms.

Wavelet transform [2] has been proposed to detect and classify various types of power quality disturbances. The wavelet analysis expands the signal in terms of wavelets, which are generated in the form of translations and dilations of a fixed function called the mother wavelet. Wavelet transformation has the ability to analyze different power quality problems simultaneously in both time and frequency domains. Gaouda et al. [4] proposed an effective wavelet multi-resolution signal decomposition method for analyzing the power quality transient events based on the standard deviation and root mean square value. Huang and Jou [5] proposed an arithmetic coding approach based on wavelet packet transform to compress the power quality disturbance data in their paper. Since noise is omnipresent in a real electrical power distribution network, Yang and Liao [6] presented a de-noising scheme for enhancing wavelet-based power quality monitoring system. In this scheme, Gaussian white noise is considered and a threshold to eliminate the noise influence is determined adaptively according to the background noise.

Gaing [2] has proposed a combined wavelet transform and probabilistic network (PNN) approach for disturbance waveform classification. In this approach, energy distribution at 13 decomposition levels of wavelet and time duration of each disturbance are taken as features of the network. A self organizing learning array system based on wavelet transform has been presented in [3] for the classification of power quality disturbances. The features are obtained by calculating the energy at each decomposition level.

Although wavelet transform has the capability to extract features from the signal in both time and frequency domain simultaneously and has been applied in the detection and classification of power quality, it exhibits some disadvantages [3] like excessive computation, sensitivity to noise level and the dependency of its accuracy on the chosen basis wavelet.

Mishra et al. [7] has proposed S-transform approach for feature extraction in disturbance waveform classification. In this approach, the features of the network are extracted from the frequency and phase contours of the S-matrix of the signal. Lee [11] has presented S-transform based intelligent based system for classification of power quality disturbance signals in which the change in energy and standard deviation calculated from the S-transform of the contour is considered as a feature to the network. Samantaray et al. [13] employed S-transform based statistical techniques for the analysis of power quality disturbances.

In this paper, Hilbert transform (HT) has been proposed for feature extraction. Hilbert transform transforms the real data sequence into an analytical signal which has a real part, that contains the original data, and an imaginary part, which contains the Hilbert transform. The imaginary part is a version of the original real sequence with a 90° phase shift. Sines are, therefore, transformed to cosines and vice versa. The Hilbert transformed series has the same amplitude and frequency content as the original real data and includes phase information that depends on the phase of the original data. The Hilbert transform is useful in calculating instantaneous attributes of a time series, especially the amplitude and frequency. The HT operator is a linear one, capable of tracking the amplitude envelope of the signal. The advantage of Hilbert transform over wavelet transform is that it avoids the requirement of testing various families of wavelets so as to identify the best one for the accurate classification. Further, the decomposition of the disturbance signals at different resolution levels is not required in the Hilbert transform, thereby reducing the memory size and computational overhead. Abdel-Galil and Sadaany [15] proposed the Hilbert transform for the analysis of flicker signal. In paper [16], the authors applied the Hilbert transform for calculating the envelope of the flicker signal. Yu et al. [17] employed Hilbert transform based spectrum analysis for the analysis of mechanical signals. Oin et al. [18] used Hilbert transform based envelope detection and instantaneous frequency estimation technique for the fault diagnosis of mechanical signals and in paper [19], the authors used Hilbert transform based spectral estimation techniques for the analysis of mechanical signals.

Classification is another major task in power quality recognition. Most of the authors have used feed forward neural networks with sigmoidal nonlinearities [12] for model development. The short coming of this network is that it takes long time for training. Also, it has no inherent ability to detect the outliers. In this paper, we propose RBF networks [8] to recognize the disturbance waveform. RBF networks takes less time for training and the distance-based activation function used in the hidden nodes gives the ability to detect the outliers during estimation. Vasillic and Kezunovic [14] proposed the Fuzzy ART neural networks for the analysis of and classification of power system faults.

Section snippets

Proposed methodology

The proposed methodology for disturbance waveform classification is based on artificial neural network. Artificial neural network approach for any application involves two stages: network development and actual usage of the network. The various stages involved in the network development are:

  • data generation,

  • feature extraction and

  • network training

Generation of the appropriate training data is an important step in the development of ANN models. A large number of training data is generated through

Feature extraction using Hilbert transform

The discrete Hilbert transform [15] is a mathematical tool used to generate an analytical signal from real signal. It is obtained by convolving the real signal x(t) with the function (1/πt) as given below:xH(t)=x(t)(1/πt)=1/πx(λ/(tλ)dλwhere λ is the shifting operator.

The analytical signal provides the information about the amplitude as well as the phase of the signal.

Since the output of the DHT is, in fact 90° phase shifted version of the original signal x(t), a complex signal (also known

Radial basis function neural networks

Radial basis function network [9] is a class of single hidden layer feedforward neural network. The architecture of RBF neural network is shown in Fig. 4. The network has an input layer, a hidden layer and an output layer. The transfer functions in the hidden layer nodes are similar to the multivariate Gaussian density functionΦj(x)=exp(||xμj||2/2σj2)where x is the input vector, μj and σj are the center and spread of the corresponding Gaussian function. Each RBF unit has a significant

Simulation results

This section presents the details of the ANN-based model developed for disturbance waveform classification. The sample power system shown in Fig. 5 was used to generate the data required to develop the neural network.

The system was simulated in Matlab/Simulink and the various power quality events such as sag, swell, transients, harmonics and voltage fluctuation along with normal signal were generated by creating various faults and through harmonic injection. Faults were created at different

Conclusion

In this paper, an attempt has been made to extract efficient features of the power quality disturbances using Hilbert transform and to classify the disturbance signal using RBF neural networks. The features extracted from the Hilbert transform are very simple and yet very effective. The RBF network takes less time for training and the classification accuracy is very high. Compared to the wavelet transform and S-transform, the Hilbert transform can be quickly calculated, so that the proposed

T. Jayasree has received her BE degree in Electronics and Communication Engineering in 1997 from Bharathidasan University, Trichy, India and M.E degree in Applied electronics in 1999 from Bharathiyar University, Coimbatore, India. From 1999 to 2005, she worked as lecturer in Noorul Islam College of Engineering, Nagercoil, India. From 2006 to June 2009, she was a lecturer in Govt. Polytechnic College, Tuticorin, India and pursuing her Ph.D. degree at Anna university, Chennai, India. Now she is

References (19)

  • D. Yu et al.

    Application of EMD method and Hilbert spectrum to the fault diagnosis of bearings

    Mechanical Systems and signal processing

    (2005)
  • Y. Qin et al.

    Research on iterated Hilbert transform and its application in mechanical fault diagnosis

    Mechanical Systems and Signal Processing

    (2008)
  • H. Lev et al.

    Application of staggered under sampling to power quality monitoring

    IEEE Trans. Power Delivery

    (2000)
  • Z.L. Gaing

    Wavelet based neural network for power disturbance recognition and classification

    IEEE Trans. Power Delivery

    (2004)
  • H. He et al.

    A self organizing learning array system for power quality classification based on wavelet transform

    IEEE Trans. Power Delivery

    (2006)
  • A.M. Gaouda et al.

    Power quality detection and classification using wavelet multiresolution signal decomposition

    IEEE Trans. Power Delivery

    (1999)
  • S.J. Huang et al.

    Application of arithmetic coding for electric power disturbance data compression with wavelet packet enhancement

    IEEE Trans. Power Syst.

    (2004)
  • H.T. Yang et al.

    A de-noising scheme for enhancing wavelet based power quality monitoring system

    IEEE Trans. Power Delivery

    (2001)
  • S. Mishra et al.

    Detection and classification of power quality disturbances using S-transform and probabilistic neural networks

    IEEE Trans. Power Delivery

    (2008)
There are more references available in the full text version of this article.

Cited by (72)

  • Hilbert filter based FPGA architecture for power quality monitoring

    2019, Measurement: Journal of the International Measurement Confederation
    Citation Excerpt :

    In this sense, a vastly unexplored alternative is the use of the Hilbert Transform (HT) for the detection of a wide range of disturbances in voltage signals. Current efforts on this field yield good results, applying not only Hilbert analysis [15,17], but also more complex neural networks [18] to classify faulty behaviors on voltage signals. In these efforts, the HT has been combined with other techniques such as the wavelet transform (WT) [15] and the empirical mode decomposition (EMD) method [16,17].

View all citing articles on Scopus

T. Jayasree has received her BE degree in Electronics and Communication Engineering in 1997 from Bharathidasan University, Trichy, India and M.E degree in Applied electronics in 1999 from Bharathiyar University, Coimbatore, India. From 1999 to 2005, she worked as lecturer in Noorul Islam College of Engineering, Nagercoil, India. From 2006 to June 2009, she was a lecturer in Govt. Polytechnic College, Tuticorin, India and pursuing her Ph.D. degree at Anna university, Chennai, India. Now she is working in Govt. Polytechic College, Nagercoil, India. Her research interests are Applications of Computational Intelligent Techniques, Power quality monitoring, Signal processing.

D. Devaraj has received his B.E and M.E degree in Electrical and Electronics Engineering and Power System Engineering in 1992 and 1994, respectively, from Madurai Kamaraj University, Madurai, India. From 1994 to 1997, he worked as a Lecturer in Arulmigu Kalasalingam College of Engineering, Krishnankoil. He received his Ph.D. degree from IIT Madras, Chennai, in 2001. He is currently a Professor and Head of the Electrical and Electronics Engineering Department in Arulmigu Kalasalingam College of Engineering, Krishnankoil, India. His research interest includes Applications of Computational Intelligent Techniques, Power System Optimization, Power System Security, Fault Diagnosis and Data Mining. He has published many journal and conference publications in subject of his interests. Besides, he is a Director-Power System Group, TIFACCORE in Network Engineering sponsored by Government of India. He is a member of Indian Society of Technical education, India.

R. Sukanesh has received her B.E degree in Electronics and Communication Engineering from Madras University, Chennai, India in 1982 and M.E degree with the specialization in communication Systems in 1985 from Bharathiyar University, Coimbatore, India. She received her PhD degree from Madurai Kamaraj University, Madurai in 1999. From 1985 to 1991, she worked as a Lecturer in Thiagaraja College of Engineering, Madurai, India. She is presently working as a professor in Thiagaraja College of Engineering, Madurai. Her research interest includes Bio-Medical signal processing, image processing and neural networks. She is a member of Bio-Medical society of India (BMESI). She also a member of Institution of Engineering, Indian Society of technical education (ISTE) and Indian Association of Bio-Medical scientists (IABMS).

View full text